Gesture Segmentation and Recognition Based on Infrared Images Using Deep Neural Networks

Computers have become integral to modern life, with traditional human-computer interaction (HCI) relying on devices like the mouse and keyboard. Hand gestures offer a natural alternative, though their variability introduces challenges. Convolutional Neural Networks (CNNs), known for their ability to learn complex patterns, are well-suited for gesture recognition. This paper presents a static hand gesture recognition method using CNNs for segmentation and classification. Data augmentation techniques such as re-scaling, rotation, and shifting were applied to address gesture variability and improve accuracy. The model was tested on a near-infrared hand gesture dataset of ten poses, capturing details even in low light. Experimental results on seven subjects show an average recognition accuracy of 95%, highlighting the method's feasibility and reliability. Computers have become integral to modern life, with traditional human-computer interaction (HCI) relying on devices like the mouse and keyboard. Hand gestures offer a natural alternative, though their variability introduces challenges. Convolutional Neural Networks (CNNs), known for their ability to learn complex patterns, are well-suited for gesture recognition. This paper presents a static hand gesture recognition method using CNNs for segmentation and classification. Data augmentation techniques such as re-scaling, rotation, and shifting were applied to address gesture variability and improve accuracy. The model was tested on a near-infrared hand gesture dataset of ten poses, capturing details even in low light. Experimental results on seven subjects show an average recognition accuracy of 95%, highlighting the method's feasibility and reliability.